论文信息 - Item Preference Parameters from Grouped Ranking Observations

Item Preference Parameters from Grouped Ranking Observations

Given a set of rating data for a set of items, determining the values of items is a matter of importance. Various probability models have been proposed to solve this task. The Plackett-Luce model is one of such models, which parametrizes the value of each item by a real valued preference parameter. In this paper, the Plackett-Luce model is generalized to cope with the grouped ranking observations such as movies or restaurants ratings. Since the maximization of the likelihood of the proposed model is computationally intractable, the lower bound of the likelihood which is easy to evaluate is derived, and the em algorithm is adopted to the maximization of the lower bound.

[1] R. A. Bradley,et al. Rank Analysis of Incomplete Block Designs: I. The Method of Paired Comparisons , 1952 .

[2] R. Plackett. The Analysis of Permutations , 1975 .

[3] M. Fligner,et al. Distance Based Ranking Models , 1986 .

[4] Gediminas Adomavicius,et al. Toward the next generation of recommender systems: a survey of the state-of-the-art and possible extensions , 2005, IEEE Transactions on Knowledge and Data Engineering.

[5] R. Duncan Luce,et al. Individual Choice Behavior , 1959 .

[6] C. L. Mallows. NON-NULL RANKING MODELS. I , 1957 .

[7] Shun-ichi Amari,et al. Information geometry of the EM and em algorithms for neural networks , 1995, Neural Networks.

[8] D. Hunter. MM algorithms for generalized Bradley-Terry models , 2003 .

[9] Chih-Jen Lin,et al. Generalized Bradley-Terry Models and Multi-Class Probability Estimates , 2006, J. Mach. Learn. Res..

[10] H. A. David,et al. The method of paired comparisons , 1966 .

[11] R. A. Bradley,et al. RANK ANALYSIS OF INCOMPLETE BLOCK DESIGNS , 1952 .

[12] Tommi S. Jaakkola,et al. Tutorial on variational approximation methods , 2000 .